Llama-2-70b-hf
Access to model meta-llama/Llama-2-70b-hf is restricted.
text → text · made by Meta Llama
Sold by
Nobody in the catalogue publishes a price for this yet.
Measured 12 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 1stof 115 | Trivia qa — Epoch AI | EM | 0.876 |
| 5thof 53 | Lambada — Epoch AI | Score | 78.9 |
| 6thof 6 | Common sense qa 2 — Epoch AI | Score | 50 |
| 9thof 136 | Bool q — Epoch AI | Score | 88.6 |
| 13thof 113 | Hella swag — Epoch AI | Overall accuracy | 0.853 |
| 17thof 88 | Bbh — Epoch AI | Average | 0.649 |
| 17thof 113 | Piqa — Epoch AI | Score | 82.8 |
| 19thof 70 | Open book qa — Epoch AI | Accuracy | 60.2 |
| 20thof 134 | Arc ai2 — Epoch AI | Challenge score | 0.783 |
| 23rdof 137 | Wino grande — Epoch AI | Accuracy | 80.2 |
| 27thof 170 | Gsm8k — Epoch AI | EM | 0.696 |
| 72ndof 217 | Mmlu — Epoch AI | EM | 0.699 |
About
Llama-2-70b-hf — a text model from Meta Llama, placed 1st of 115 on Trivia qa — Epoch AI.
It takes text and returns text. It was published in July 2023. The catalogue files it under chat. Its weights are published under the llama2 licence, which attaches conditions the plain open licences do not. It has been measured on 12 boards, and stands best at 1st of 115 on Trivia qa — Epoch AI.
Every current figure
- Maker
- Meta Llama
- Register
- model
- Takes
- text
- Returns
- text
- Published
- July 2023
- Parameters
- 69 billion · read from its own weights
- Licence
- llama2
- Boards
- 12
- Best place
- 1st of 115 — Trivia qa — Epoch AI
Known as 1 name
meta-llama/Llama-2-70b-hf